Move the wait, keep the dependency

The load is already issued. The agent changes what the consumer does before waiting.

Before

Wait immediately

TMA engine
Load operand tileLoad complete
Consumer warp
waitIndependent workConsume tile
time

After

Do independent work before waiting: same work, earlier finish

TMA engine
Load operand tileLoad complete
Consumer warp
Independent workwaitConsume tile
time

Same time scale in both schedules; durations are illustrative.